PBB-VPLS Doubt

[mpresswa]

Hi Experts,

I was recently going through the PBB-VPLS configuration and have one query:

• One of the main reason for PBB-VPLS is used so that the MAC address table size is reduced on the PE, but in the I-domain we still learn the customer MACs and pass that information to all the remote PEs along with the PBB-encapsulation.So still the PEs will keep learning more and more MACs as we increase the number of I-domain instances. How are we solving the problem of MAC explosion in the I-domain ? The MAC explosion problem only gets solved in the B-domain.

I haven't yet configured PBB-VPLS due to hardware limitations. 

Thanks,

Muffadal

[xthuijs · Accepted solution]

Hi Muffadal,

the way to control MAC in the isid is by using the mac learning limits and associated configurable actions. When the limit exceeds we can just log, or we can flood or we can stop the forwarding.

Example:

RP/0/RSP0/CPU0:Ironman(config)#l2vpn
RP/0/RSP0/CPU0:Ironman(config-l2vpn)#bridge group PBB
RP/0/RSP0/CPU0:Ironman(config-l2vpn-bg)#bridge-domain PBB-1
RP/0/RSP0/CPU0:Ironman(config-l2vpn-bg-bd)#mac limit action ?
  flood     Stop learning but continue flooding
  no-flood  Stop learning and stop flooding
  shutdown  Stop forwarding
RP/0/RSP0/CPU0:Ironman(config-l2vpn-bg-bd)#mac limit notification ?
  both  Generate syslog message and SNMP trap
  none  No notification
  trap  Generate SNMP trap
 

cheers!

xander

[mpresswa]

First of all thanks xander for the response.

So the biggest gain while using PBB-VPLS will be reduction in the number of remote PE pseudowires via B-domain and multiple i-domains using the same overlay network i.e B-domain.

MAC learning control is implemented the same way as in the plain VPLS setup.

Regards,

Muffadal

[xthuijs · Accepted solution]

yeah the same way that MPLS reduces the routing scale needs in teh core, PBB-VPLS does that for MAC. It isolates edge vs core in terms of mac learning and makes it a lot more controlable.

Same with the reduction on the number of PW's that are needed between edge nodes. It is truly a simplified design.

cheers!

xander

[mpresswa]

Thanks xander for the clarifications. It is indeed a simplified design from deployment point of view. 

Regards,

Muffadal